PNNL

Abstract

The development and testing of improved alarm algorithms is an ongoing priority of the Radiation Portal Monitor Project (RPMP) at Pacific Northwest National Laboratory (PNNL). Improved algorithms have the potential to reduce the impediments that radiation screening presents to the flow of commerce, without affecting the detection sensitivity to sources of interest. However, assessing alarm-algorithm performance involves careful calculation of detection probabilities and nuisance/false alarm rates for any algorithm that may be used in the field. To establish statistical confidence, such a task requires a large amount of data from drive-through (or “dynamic”) scenarios both with, and without, radioactive sources of interest present; but obtaining actual field data to meet this need is not feasible. Instead, an “injection-study” procedure is being used to approximate how the profiles of actual drive-through commercial data would change with the presence of sources of interest. This procedure adds net-counts from a pre-defined set of simulated sources to raw, gross-count drive-through data randomly selected from archived cargo data collected from deployed radiation portal monitors (RPMs). (PIET-43741-TM-480)